Composite moving averages in forecasting; fairly accurate-- even with fairly random actual sales fig

[socraticquest]

Hello,

I use composite moving averages in order to forecast weekly sales for a
small-business. I've taken a few years of 200 previous weeks sales of
data and attempt to project the next weeks sales.

Considering that the sales figures (when charted) are fairly random,
I'm amazed that the forecasted projections come so close to the actual
weekly outcomes. In attempting to forecast forty possible outcomes, the
actual outcomes are often plus or minus one or two points away from the
forecasted outcomes.

Has anybody noticed "narrow error ratios"...so often? Any
recommendations for software that best deals deal with error-ratios in
order to yield even more accurate projections?

Thank-you

 

[dave]

S.

The anaytical aproach to your problem us called BOX-JENKINS . The
method of BOX-JENKINS also known as ARIMA modelling dtermines BOTH the
number of periods to be used AND the specific weights that should be
applied ( as compared to EQUAL WEIGHTS ) .

We write and sell software to so this ...Please visit
http://www.autobox.com the home page for AUTOBOX and FREEFORE ( which
has a FreeWare version ..http://www.autobox.com/Freefore-over.htm

If I can help, please call me at 215-675-0652

Dave Reilly
Automatic Forecasting Systems.

P.S. Your narrow confidence intervals is a nice complaint ..bur
obviously might be made even narrower with more appropriate modelling.

 

[steve]

Has anybody noticed "narrow error ratios"...so often? Any
recommendations for software that best deals deal with error-ratios in
order to yield even more accurate projections?

If you send me a copy of the data I'll run it through he forecast in
my software and send you the results. I've no idea how accurate it be
but I'm willing to give it go.

 

[FTR]

S.

The anaytical aproach to your problem us called BOX-JENKINS . The
method of BOX-JENKINS also known as ARIMA modelling dtermines BOTH the
number of periods to be used AND the specific weights that should be
applied ( as compared to EQUAL WEIGHTS ) .

We write and sell software to so this ...Please visit
http://www.autobox.com the home page for AUTOBOX and FREEFORE ( which
has a FreeWare version ..http://www.autobox.com/Freefore-over.htm

You mean the demo version ? I don't see a freeware.

 

[dave]

This is a full functioning version ....

But only forecasts out 1 period for NEW DATA for those that don't
purchase the PROFESSIONAL VERSION ,,,, but it will give you the model
and the parameters which you can then take into excel for a multiple
period out forecast,...

DAVE ..

p.s. If you want to send your data to us , we will generate a
multi-period forecast for you in order to demonstrate the real
functionality of our software.

dave r